Exploration Of Immunotherapy Biomarkers And Application Of Artificial Intelligence Technology In Patients With Non-small Cell Lung Cancer

1 BackgroundIn the world,cancer is the leading cause of human death,and it ranks first or second deaths in 112 countries.And lung cancer had the highest mortality(1.8 million,18%).Lung cancer includes two major pathological types:non-small cell lung cancer(NSCLC)accounts for about 85%,and small cell lung cancer(SCLC)accounts for 15%.Traditional managements,including surgery,chemotherapy,radiotherapy and targeted therapy,are not completely satisfying with 5-year survival rate remaining 16%,which is still very low.With the discovery of immune checkpoints,such as programmed death1(PD-1)and its ligand(PD-L1),cytotoxic T lymphocyte associated atigen-4(CTLA4),the treatment of cancer have completely revolutionized with the supporting of immune checkpoint inhibitors(ICIs).Immunotherapy provided unprecedented level of survival benefit to cancer patients,including patients who were incurable with advanced cancer,and revolutionized the oncology,including lung cancer.Although the feasibility of PD-1 inhibitor monotherapy as the first-line treatment for patients with PD-L1-positive tumors has been confirmed again,it has not changed the situation that only patients with high PD-L1 expression are recommended for firstline treatment.Besides,only a few of NSCLC patients benefit from immune checkpoint inhibitor therapy.In general,the response rate of immunotherapy is about 20%-30%,while in patients with EGFR mutation or ALK rearrangement after receiving immunotherapy,the ORR is only 4%.For EGFR wild-type and ALK-negative NSCLC patients,ORR is also only about 23%.Even for the NSCLC patients with PD-L1 expression over 50%,the response rate is only about 40%.In addition,ICIs treatment can also induce hyperprogression(6%-29%)and pseudoprogression(4.5%).This results in a greater economic and physical burden for patients who cannot benefit from immunotherapy.Therefore,it is one of the hot topics to find biomarkers that can identify potential responders to immunotherapy,so as to achieve the purpose of precise treatment of lung cancer,and it is also the purpose of this paper.This paper will discuss this issue from the following three aspects.2 The genomic landscape of patients with lung squamous cell carcinoma and its relationship with the immune landscape2.1 ObjectionIn this study,we intended to explore the correlation of TMB,PD-L1 expression and CD8+tumor-infiltrating lymphocytes(TIL)with genomic status and disease survival based on lung cancer pathological tissue specimens.2.2 Methods(1)We retrospectively collected 180 patients with lung squamous cell carcinoma who underwent lung surgery(lobectomy or pneumonectomy)and were histologically confirmed in Shanghai Pulmonary Hospital from 2013 to 2016,all of the patients did not receive immunotherapy;(2)The expression of PD-L1 was detected by E1L3N antibody,and the density of CD8+TILs was detected by mouse anti-human CD8 monoclonal antibody(M7103,clone C8144B,DAKO);(3)Genomic DNA was extracted from all included samples using Qiagen DNeasy FFPE DNA kit(Qiagen,Hilden,Germany),and a DNA library was constructed,and then the captured DNA fragments were sequenced using an Illumina Novaseq 6000 sequencing system.TMB is defined as the number of somatic coding mutations containing single-base substitutions and deletions detected in the genome per megabase,considering only non-synonymous mutations,excluding intronic and synonymous mutations that do not affect the amino acid sequence.2.3 Results(1)We detected 180 samples using targeted sequencing,and important somatic mutations were consistent with previous studies,only PTCH1 gene expression was significantly different between CD8+TIL+and CD8+TIL-patients;(2)Survival analysis showed that only high PD-L1 expression and CD8+tumorinfiltrating lymphocyte infiltration were significantly associated with longer survival time;subgroup analysis results showed that the association between PD-L1 expression and CD8+TILs and survival was significant only in the stage Ⅰ/Ⅱ lung squamous cell carcinoma cohort.2.4 ConclusionThe mutation frequency of PTCH1 gene was correlated with the density of CD8+TIL.Both high CD8+TIL and high PD-L1 expression were independent prognostic parameters of DFS.These biomarkers provide new ideas for the precision treatment of early LUSC.3 PD-L1 features of NSCLC patients’ pathological tissues were explored with artificial intelligence technology based on IHC images3.1 ObjectionThis study aims to combine artificial intelligence technology and build an artificial intelligence model based on u-net++algorithm 86 of cascade training,so as to be able to read and divide pathological slide images of PD-L1 immunohistochemical staining and complete automatic detection of PD-L1 positive lung cancer cells,so as to assist pathologists in clinical work and reduce their burden.3.2 Methods(1)We collected the whole-slides of patients with non-small cell lung cancer who received immunotherapy in Shanghai Pulmonary Hospital from February 2016 to July 2021.All of the slides contained PD-L1 immunohistochemical staining(the antibody includes 22C3 and E1L3N);(2)Model training process:scan to obtain a pathological slide image set-obtain the coordinates of the region of interest-crop the region of interest into a small image set-manually label PD-L1 positive tumor cells and PD-L1 on the small block Negative tumor cells—apply U-net++algorithm to train the model--obtain the segmentation results of PD-L1 positive lung cancer cells identified by the model--evaluate the segmentation results of the model.3.3 Results(1)Complete the construction and testing of the U-net++model based on cascade training in the PyTorch packages;(2)Obtain the PD-L1 positive lung cancer cell density map segmented by the model and the calculated TPS score;(3)The image results segmented by this model are compared with the segmentation effects of existing open-source medical image software and the TPS diagnosed by pathology experts.The results show that the model constructed in this study has high accuracy and consistency.3.4 ConclusionWe built a U-net++-based on cascade training net++structure model,can be done reliably panoramic pathological slides of immunohistochemical staining in PDautomated testing using the model of L1 positive lung cancer cells to provide the high resolution of PD-L1 positive segmentation map the depth of the extracted features of lung cancer cells have good ability to assess the prognosis.4.The exploration of immune indexes based on peripheral blood samples and the efficacy of NSCLC immunotherapy4.1 ObjectionIn this part,we intended to explore the relationship between hematological markers and clinical benefit of patients with advanced NSCLC receiving immunotherapy based on peripheral blood biological samples.4.2Methods(1)We retrospectively collected 87 patients with advanced NSCLC who received at least 1 dose of first-line ICI with or without anti-angiogenesis or chemotherapy in Shanghai Pulmonary Hospital from January 2018 to June 2019.The follow-up period ended in 2022.March;(2)The clinical characteristics and peripheral blood test indexes of patients were collected from electronic medical records,including age,gender,Eastern Cooperative Oncology Group Performance Status(ECOG PS),smoking history,tumor proportion score of PD-L1 expression(TPS).Obtain baseline peripheral blood parameters,including hemoglobin(Hb);red blood cell(RBC);albumin(ALB)and so on.TPS was defined as the percentage of PD-L1-positive tumor cells in all tumor cells,and PD-L1 positive expression was defined as TPS≥1%;(3)Data analysis:Continuous variables are expressed as median and range,and categorical variables are expressed as counts.The comparative analysis of continuous and categorical variables was performed using two independent samples t-test,Pearson’s chi-square test or Fisher’s exact test,respectively.Quartile analysis was used to determine the optimal cut-off values for clinical indicators.PFS and OS were estimated using the Kaplan-Meier.Univariate and multivariate analyses of the effects of each indicator on survival were performed using Cox proportional hazards models,all statistical tests were bilateral,and the significance threshold was P=0.05.All statistics was analyzed by SPSS Version 25(SPSS Inc.,Munich,Germany).4.2 ResultsThe results of this part showed that in patients with advanced NSCLC receiving first-line ICIs treatment,baseline albumin(ALB)was an independent predictor.Compared with patients with low ALB(≤3 8g/L),patients with high ALB(>3 8g/L)had significantly prolonged PFS and OS.4.2 ConclusionIn advanced NSCLC patients receiving first-line ICI treatment,baseline serum ALB level was an independent predictor of prognosis in advanced NSCLC patients receiving ICIs treatment.High ALB(>38g/L)is a protective factor for longer PFS and OS in NSCLC patients.The K-M graph shows that NSCLC patients with high AEC levels have a poor TREND in PFS and OS,but the difference analysis results are not significant,and it may be necessary to find a new optimal threshold for further analysis.